Despite the existing medical infrastructure being limited in terms of interoperability, the amount of medical multimedia transferred over the network and shared through various channels increases rapidly. In search of consultations with colleagues, medical professionals with the consent of their patients, usually exchange medical multimedia, mainly in the form of images, by using standard instant messaging services which utilize lossy compression algorithms. That consultation paradigm can easily lead to losses in image representation that can be misinterpreted and lead to the wrong diagnosis. This paper presents MIDOM—Medical Imaging and Diagnostics on the Move, a DICOM-based medical image communication system enhanced with a couple of variants of our previously developed custom lossless Classification and Blending Predictor Coder (CBPC) compression method. The system generally exploits the idea that end devices used by the general population and medical professionals alike are satisfactorily performant and energy-efficient, up to a point to support custom and complex compression methods successfully. The system has been implemented and appropriately integrated with Orthanc, a lightweight DICOM server, and a medical images storing PACS server. We benchmarked the system thoroughly with five real-world anonymized medical image sets in terms of compression ratios and latency reduction, aiming to simulate scenarios in which the availability of the medical services might be hardly reachable or in other ways limited. The results clearly show that our system enhanced with the compression methods in the question pays off in nearly every testing scenario by lowering the network latency to at least 60% of the latency required to send raw and uncompressed image sets and 25% in the best-case, while maintaining the perfect reconstruction of medical images and, thus, providing a more suitable environment for healthcare applications.

Sudden cardiac death (SCD) remains a major clinical challenge, with implantable cardioverter-defibrillators (ICDs) serving as the primary preventive intervention. Current patient selection guidelines rely on limited and imperfect risk markers. This study explores the potential of machine learning (ML) models to improve SCD risk prediction using tabular clinical data that include features derived from medical sensing devices such as electrocardiograms (ECGs) and ICDs. Several ML models, including tree-based models, Naive Bayes (NB), logistic regression (LR), and voting classifiers (VC), were trained on demographic, clinical, laboratory, and device-derived variables from patients who underwent ICD implantation at a Croatian tertiary center. The target variable was the activation of the ICD device (appropriate or inappropriate/missed), serving as a surrogate for high-risk SCD detection. Models were optimized for the F2-score to prioritize high-risk patient detection, and interpretability was achieved with post hoc SHAP value analysis, which confirmed known and revealed additional potential SCD predictors. The random forest (RF) model achieved the highest F2-score (F2-score 0.74, AUC-ROC 0.73), demonstrating a recall of 97.30% and meeting the primary objective of high true positive detection, while the VC classifier achieved the highest overall discrimination (F2-score 0.71, AUC-ROC 0.76). The predictive performance of multiple ML models, particularly the high recall they achieved, demonstrates the promising potential of ML to refine ICD patient selection.

Background: Sudden cardiac death (SCD) remains a major cause of cardiovascular mortality. Implantable cardioverter-defibrillators (ICDs) reduce arrhythmic mortality, but current selection based largely on left ventricular ejection fraction (LVEF) lacks precision. Many patients undergo device implantation without ever receiving therapy, while others at risk remain unprotected. Interpretable machine learning (ML) can integrate diverse clinical variables and refine patient selection while maintaining transparency in clinical reasoning. Methods: We retrospectively analyzed 607 patients who underwent ICD or CRT-D implantation at a Croatian tertiary care center. Baseline demographic, clinical, echocardiographic, laboratory, and device-related variables were collected. Patients were followed through routine device interrogations, with appropriate ICD activation serving as a surrogate for SCD prevention. A logistic regression (LR) model was trained to predict appropriate device activation. Results: LR model demonstrated strong predictive ability (AUC-ROC 0.74, sensitivity 86.50%). Significant predictors included ventricular tachycardia (VT) burden, sustained VT, longer follow-up, and secondary prevention. The combination of furosemide and spironolactone therapy was linked to lower predicted SCD risk. Conclusions: ML applied to routinely collected data can support risk stratification in SCD and complement existing guideline criteria by reinforcing known predictors and uncovering novel associations.

Heterogeneous computing has emerged as a key paradigm for improving performance and energy efficiency. While fully customized accelerators achieve the best results in the applications for which they are designed, in terms of performance and energy efficiency, programmable accelerators strike a balance between these gains and applicability to multiple application domains. However, such systems are inherently hard to program. They are often programmed using imperative languages through low-level programming interfaces, requiring deep knowledge of the underlying architectural details. For domain scientists who could make the most use of the processing power available in such systems, the lack of abstraction in the programming model renders the learning curves unnecessarily steep. In this paper, we address the problem of programmability for heterogeneous embedded systems with programmable accelerators by systematically extending the domain-specific programming language Rise and its compiler Shine for an exemplary edge computing heterogeneous platform (PULP). Ultimately, Rise enables the programmer to express a problem in terms of what is being computed, while Shine optimizes the expression and compiles it to the language native to the target platform. We evaluated our approach against the native programming model (C) of our test platform as well as against other prominent solutions targeting embedded heterogeneous hardware. Our results show performance comparable to competing approaches while offering a cleaner and more concise programming model.

This paper presents the improved version of our cool Cracker cluster (cCc), a heterogeneous distributed system for parallel and energy-efficient bcrypt password hash computation. The cluster consists of up to 8 computational units (nodes) with different performances measured in bcrypt hash computations per second [H/s]. In the cluster, nodes are low-power heterogeneous embedded systems with programmable logic containing specialized hash computation accelerators. In the experiments, we used a combination of Xilinx Zynq-series SoC boards and ZTEX 1.15y board which was initially used as a bitcoin miner. Zynq based nodes use the improved version of our custom bcrypt accelerator, which executes the most costly parts of the bcrypt hash computation in programmable logic. The cluster was formed around the famous open-source password cracking software package John the Ripper (abbr. JtR). On the communication layer, we used Message Passing Interface (MPI)library with a standard Ethernet network connecting the nodes. To mitigate the different performances among the cluster nodes and to balance the load, we developed and implemented password candidate distribution scheme based on the passwords' probability distribution, i.e. the order of appearance in the dictionary. We tested individual nodes and the cluster as a whole, trying different combinations of nodes and evaluating our distribution scheme for password candidates. We also compared our cluster with various GPU implementations in terms of performance, energy-efficiency, and price-efficiency. We show that our solution outperforms other platforms such as high-end GPUs, by a factor of at least 3 in terms of energy-efficiency and thus producing less overall cost of password attack than other platforms. In terms of the total operational costs, our cluster pays off after 4500 cracked passwords for a bcrypt hash with cost parameter 12, which makes it more appealing for real-world password-based system attacks. We also demonstrate the scalability of our cCc cluster.

Beat the Chasers is a popular UK-originating TV quiz show that premiered in Croatia in 2023. On the show, a contestant challenges a team of up to five chasers with respect to the offers provided by the production. Each offer balances risk and reward, varying in prize money, time advantage, and the number of chasers. In this paper, we first present the dataset obtained by extracting data from the publicly broadcast episodes of Beat the Chasers in Croatia. We then apply various machine-learning models with the goals of predicting (1) which offer a contestant is most likely to select and (2) the game’s outcome. The best-case results suggest that we can successfully do both by reaching an F1-score of $73.6\%$ for the selected offer prediction and $84.6\%$ for the game’s outcome prediction. Regarding the feature importance analysis, we identified the contestant’s hometown size, NUTS 2 region, age group, and gender as the most relevant features in the case of the selected offer prediction. As for the outcome prediction, the game-specific features emerged as the most important, namely, the cash builder result, the selected number of chasers, and the chasers’ time in the selected offer.